We propose to investigate methods for inference and model criticism for relative risk regression models. We will first show how the likelihood of case-control and survival studies can be viewed as a realization of a multinomial random variable. This will suggest why there are difficulties in inference for general relative risk functions. These difficulties in inference for general reactive risk functions. These difficulties carry over to diagnostics where score based on the score to odds ratio estimation and then to investigate extensions to more general problems. It is anticipated that the cohort studies of humans and animals described in the other projects will be enhanced by having the most modern methods available for analysis of their data.